Table 5 Comparison of experimental data of different algorithms on the thermal infrared dataset.
From: An improved algorithm for infrared road object recognition in UAV perspective
Model | Model parameters (M) | P (%) | R (%) | mAP 0.5 (%) | F1 (%) |
|---|---|---|---|---|---|
Faster-RCNN | 136.8 | 87.5 | 88.1 | 87.1 | 87.8 |
YOLOv5 | 46.1 | 92.2 | 87.6 | 91.1 | 89.8 |
YOLOv6 | 58.3 | 90.4 | 89.1 | 91.3 | 89.7 |
YOLOv7 | 37.2 | 90.2 | 91.1 | 91.5 | 90.6 |
YOLOv8 | 43.6 | 90.4 | 87.0 | 90.6 | 88.7 |
YOLOv11 | 25.3 | 91.7 | 87.8 | 91.8 | 89.7 |
YOLOv12 | 26.4 | 92.1 | 89.4 | 92.6 | 90.7 |
RetinaNet | 36.4 | 90.2 | 90.7 | 91.2 | 90.4 |
Ours | 46.2 | 94.5 | 92.9 | 95.7 | 93.7 |
